Eur J Clin Pharmacol (2009) 65:705713

DOI 10.1007/s00228-009-0637-4

PHARMACOKINETICS AND DISPOSITION

Individualised dosing of amikacin in neonates:

a pharmacokinetic/pharmacodynamic analysis
Catherine M. T. Sherwin & Sofia Svahn &
Antje Van Der Linden & Roland S. Broadbent &
Natalie J. Medlicott & David M. Reith

Received: 9 January 2009 / Accepted: 11 February 2009 / Published online: 21 March 2009
# Springer-Verlag 2009

Abstract 24 to 41 weeks. The final pharmacokinetic model analysed

Purpose To examine the pharmacokinetics of amikacin and 358 amikacin concentrations. All neonates were >72 hours
its pharmacokinetic pharmacodynamic (PKPD) relationship postnatal age. Simulations were performed to develop a
in neonates. To develop an alternative dosing strategy for new dosing strategy.
amikacin in neonates. Results The final covariate model was clearance=0.23
Methods A population PKPD analysis was performed using (current weight/2)0.691 (postmenstrual age/40)3.23 and
data collected from 80 neonates with gestational ages from volume of distribution=0.957(current weight/2)0.89. Fol-
lowing the logistic regression analysis of treatment failure,
C. M. T. Sherwin (*)
new amikacin target concentrations were estimated and
Division of Clinical Pharmacology,
Cincinnati Childrens Hospital Medical Center, used in development of an alternative dosing strategy.
3333 Burnet Avenue, MLC 6018, Conclusion Simulation of a new dosing regimen yielded
Cincinnati, OH 45229-3039, USA the following recommendations: 15 mg/kg at 36-h intervals,
e-mail: catherine.sherwin@cchmc.org 14 mg/kg at 24-h intervals and 15 mg/kg at 24-h intervals
D. M. Reith for neonates 28 weeks, 2936 weeks and 37 weeks
Department of Womens and Childrens Health, postmenstrual age respectively.
Dunedin School of Medicine, University of Otago,
P.O. Box 913, Dunedin, New Zealand Keywords Amikacin . Neonates .
e-mail: david.reith@stonebow.otago.ac.nz
Population pharmacokinetics . Pharmacodynamics
S. Svahn
University of Uppsala,
Uppsala, Sweden
Introduction
e-mail: svahn.sofia@gmail.com

A. Van Der Linden Bacterial sepsis is common in neonates and particularly in

Department of Pathology, Dunedin School of Medicine, preterm neonates because of their immature immune system
Dunedin, New Zealand
[1]. Amikacin is a semi-synthetic derivative of kanamycin
e-mail: antje.vanderlinden@otagodhb.govt.nz
which has broad in vitro and in vivo activity against most
R. S. Broadbent aerobic Gram-negative bacteria and is effective against
Department of Womens and Childrens Health, some strains of bacteria that are resistant to other amino-
Dunedin School of Medicine, University of Otago,
glycosides [2, 3]. Amikacin is indicated in the short-term
Dunedin, New Zealand
e-mail: roland.broadbent@otago.ac.nz treatment of serious infections caused by strains of
Pseudomonas sp., E. coli, Klebsiella pneumoniae, Serratia
N. J. Medlicott sp. and Staphylococcus species [4]. Specifically in the
New Zealands National School of Pharmacy,
neonatal population, amikacin is also used to treat late-
University of Otago,
Dunedin, New Zealand onset sepsis caused by coagulase-negative Staphylococcus
e-mail: natalie.medlicott@otago.ac.nz species. For Gram-negative sepsis, it is recommended that
706 Eur J Clin Pharmacol (2009) 65:705713

the amikacin peak serum concentration should be higher culture-proven sepsis, sex, serum creatinine (mol/L), max-
[maximum serum concentration (Cmax)/minimum inhibitory imum C-reactive protein (CRP) (mg/L) during treatment,
concentration (MIC) ratio greater than 8:1] in the first to Apgar score at 1 min and Apgar score at 5 min. Post-menstrual
second days of treatment [57]. However, these recom- age was calculated as gestational age plus post-natal age at
mendations have not been based upon studies in neonates, birth in completed weeks.
for which there is a paucity of data with regard to amikacin The standard NICU protocol for suspected sepsis
pharmacokinetic-pharmacodynamic (PKPD) relationships. required a blood culture, full blood count and CRP to be
Several studies have demonstrated that conventional performed. Amikacin was discontinued after 48 h unless
paediatric dosing regimens are not well adapted to neonates neonates were confirmed septic or remained clinically
and result in inconsistent serum concentrations, particularly in unwell. Neonates with culture-proven sepsis, positive
extremely low-birth-weight neonates [5]. Peak concentrations microbial infection confirmed by cerebral spinal-fluid tap,
are often lower than expected for optimal bactericidal effect, clinical unresolved signs or a persistent maximum CRP >
and trough concentrations are often too high to be considered 40 mg/L were classified as septic. For those neonates with
safe [5, 8]. The concentrations related to toxicity in adults culture-proven sepsis, treatment failure (TF) was defined as
have been extrapolated to neonates, but there is little a repeated infection caused by the same bacteria within
evidence indicating the specific toxic concentrations for 5 days and/or a treatment switch to another antimicrobial,
amikacin in neonates [9]. Given the inconsistency in the usually vancomycin, due to no apparent clinical improve-
literature and the limited evaluation of extended interval ment with amikacin treatment.
dosing with amikacin in neonates there is a need to design a
dosing regimen adapted to extremely premature neonates. Amikacin therapy and sampling
However, before a new dosing regimen can be developed,
the pharmacokinetics (PK) of amikacin need to be further Amikacin was administered as an intravenous infusion over
explored, in particular, the influence of covariates. 30 min using a syringe driver piggybacked to a Baxter pump
The present study was undertaken to investigate the delivering a continuous IV infusion. This was followed with
pharmacokinetics and pharmacodynamics of amikacin in 1 mL normal (0.9%) saline given over 30 min. Amikacin dose
neonates, and the influence of covariates upon these in the first week after birth was based on current weight and
parameters. This information was used to develop an gestational age. Neonates 27 and 2730 weeks gestational
optimal dosing strategy for neonates. age received 18 mg/kg at 48- and 36-h intervals respectively.
Neonates 31 to 33 weeks gestational age received 16 mg/kg
at 36-h intervals. Neonates 34 weeks gestational age
Methods received 15 mg/kg at 24-h intervals. From 1 week of age,
neonates were given an initial dose of 15 mg/kg and were
Patients then dosed based on gestational age as described above.

A retrospective chart review was performed that included all Amikacin and creatinine assays
neonates treated with amikacin in the neonatal intensive care
unit (NICU) at Dunedin Hospital from 1 October 2003 to 31 Amikacin peak serum concentrations were measured
January 2007. Amikacin was commenced for treatment of 60 min after the start of the IV infusion, and trough
suspected or culture-proven late-onset neonatal sepsis from concentrations were measured at 24 h. Amikacin concen-
3 days post-natal age. As per the Dunedin Hospital NICU trations were measured using a fluorescence polarisation
protocol, neonates <72 h post-natal age are treated with immunoassay with an Abbott TDx kit (Abbott Park, IL).
gentamicin and were not included in this study. Data were Serum creatinine levels were obtained within 24 h of
collected for 82 neonates who had at least one therapeutic amikacin levels and were assayed by the Jaffe alkaline
amikacin concentration recorded. Two patients were excluded picrate method (Roche Diagnostics, Indianapolis, IN).
from the review due to incomplete data. Three amikacin CRP was determined using a Roche Tinaquant, immu-
concentrations were also excluded from the PK analysis, two noturbidimetric method (Roche Diagnostics, Indianapolis,
due to intravenous leakage during administration and one was IN). Amikacin Etests (ABBiodisk, Solna, Sweden)
considered an outlier due to an unexplainable, exceptionally were performed on the isolates of bacteria cultured from
high concentration. The study protocol was reviewed by the 26 septic neonates to determine amikacin MIC. Etests
Otago Ethics Committee and considered an audit. were not performed on streptococci or enterococci
The data obtained from the review of medical charts isolates, as these are intrinsically resistant. All analyses
included gestational age (weeks), post-natal age (days), birth were carried out at the Otago Diagnostic Laboratory at
weight (kg), current weight at treatment (kg), presence of Dunedin Hospital.
Eur J Clin Pharmacol (2009) 65:705713 707

Pharmacokinetic modelling Pharmacodynamic analysis

The population PK analysis of amikacin used NON- The pharmacodynamic (PD) analysis was undertaken using
MEM version 5, release 1.1 (non-linear mixed effect a logistic regression technique for making predictions when
modelling; GloboMax, Hanover, MD, USA). A mixed the dependent variable (Y) was dichotomous rather than
effects one-compartment, first-order elimination model continuous, and the independent variables were continuous
was developed. The terms for variability in the model or discrete. The outcome of Y allowed for only two possible
included between-subject variability (BSV), between- values. These were represented by one, a success, and
occasion variability (BOV) and residual variability. An zero, a failure. The mean of the dichotomous random
omega block was used in the model to account for the variable (p) was the proportion of times Y=1 or treatment
high correlation between clearance and volume of was successful. The logistic function did not yield a value
distribution. Linear and power error models were that was either negative or greater than one, and it was
explored and the random-effect error models included restricted to estimating the value of p to the required range.
BSV, BOV and residual variability. The final BSV and Modelling p with the logistic function was equivalent to
BOV error models used an exponential error model. The fitting a linear regression model where the continuous
residual (unexplained) variability was evaluated using a outcome Y was replaced by the logarithm of the odds of
combined proportional (RUV-CV) and additive (RUV- success [11].
SD) error model. The PD analysis used treatment failure as the outcome
Plots were generated to screen for relationships variable and included all the neonates with culture-proven
between the estimates of the parameters from the base sepsis (n = 26). The logistic regression aimed to predict the
model and potential covariates. The independent varia- odds ratios for patients with TF or those likely to meet the
bles (covariates) included gestational age (weeks), post- TF criteria based on the collected clinical data. The effect
natal age (days), current weight, sepsis, serum creatinine, size of the exposure variables was expressed as odds ratio
sex, Apgar scores and post-menstrual age. A further (95% CI). Relative risk was calculated to determine the risk
exploratory analysis of plots was also undertaken by of treatment failure occurring dependent on the amikacin
plotting post-hoc eta predictions for clearance and volume therapeutic peak concentration and MIC.
of distribution against the biologically plausible covariates.
For inclusion of covariates, an additive and multiplicative Simulation of current and proposed dosing regimens
approach was used initially in a univariate analysis. A
forward and backward stepwise regression method was The final covariate PK model was used to simulate serum
used to determine the best covariate model. A statistically concentrations during the current dosing regimen. Then an
significant improvement in the fit of the model was based alternative dosing regimen was explored using simulations
on minimisation of the objective function value (OFV) and with the same model. A nonparametric dataset of covariate
hypothesis-testing using the 2 distribution. The final distributions was used to represent 1,000 virtual patients
model was determined by investigating significant mini- using MATLAB (student version 7.1, The Mathworks,
misation of the OFV and visual inspection of diagnostic 2005). The covariates of interest were post-menstrual age
plots. After identification of probable significant cova- (weeks) and current weight (kg), and these ranged from
riates, linear and power models were investigated. A 24.7 to 44.1 and 0.45 to 4.43 respectively, based on a
power model was chosen as it provided a greater reduction dataset that comprised 719 individual neonates. The
in the OFV than linear models with the same number of simulation modelled an intravenous infusion of amikacin
model parameters. given over 30 min for two doses. The new dosing regimen
was proposed to achieve treatment success as defined by
Bootstrap Cmax in the range 2435 mg/L and area under the
concentration-time curve over 24 h (AUC24) in the range
The final PK model was validated by using the bootstrap 130590 mgh/L. An extended interval dosing regimen was
approach. The sampling strategy used a non-replacement developed in line with current literature recommendations
technique [10]. A bootstrap was performed on the final for dosing aminoglycosides in neonates [8, 12, 13].
refined covariate PK model, which was based on the whole
population. The bootstrap repeatedly fitted the data to the
final model and generated 1,000 replicates. The percentage Results
of runs where the covariate effects were determined to be
significant was calculated, and each model parameter was Eighty neonatal patients were included in the final study
reported as median and 95% CI. for which there were 358 amikacin peak and trough
708 Eur J Clin Pharmacol (2009) 65:705713

concentrations. In this population, there were 46 (57.5%) with BSV estimated at 24.4% (5.83 SE %) for CL and
extremely premature neonates, with a gestational age of 4.52% (2.61 SE %) for V. For residual variability, the
<28 weeks, and 44 (55%) of these neonates were coefficient of variability (RUV-CV) was estimated at 13.7
extremely low-birth-weight infants (<1,000 g) (Table 1). (2.46 CV %) and standard deviation (RUV-SD) was 2.34
There were 98 treatment episodes including 35 (35.7%) 1.18 mg/L. The model was allowed to determine the
confirmed septic episodes from 26 (32.5%) neonates. Four allometric scale most appropriate for the current weight,
neonates each received treatment for two separate episodes resulting in a power function of 0.691 being used. Current
of culture-proven sepsis, and one patient had three treatment weight had the most significant effect on the model. The
episodes for confirmed sepsis. All episodes of culture-proven final covariate model was chosen as it produced the most
sepsis were from blood cultures, with no confirmed episodes significant minimisation of OFV ( 330.4) and reduced the
diagnosed from urine or cerebral spinal-fluid tap cultures. BSV and residual variability. Visual inspection of diagnos-
The clinical diagnoses included pneumonia (3), bacterial tic plots was also used to confirm selection of the final
meningitis (2), congenital sepsis (3) and suspected necrotis- covariate model (Fig. 1). The final covariate PK model is
ing enterocolitis (3). Seven neonates in the study population shown in equation 1.
died, five within 1 week of birth, with four of those from
overwhelming septicaemia. Twenty-one (21.4%) septic CL q1
CWT=2 q3
PMA=40 q4 L=h
1
episodes in 12 individual patients met the predefined criteria V q 2
CWT=2 q5 L
for treatment failure.
The majority of neonatal sepsis was caused by Staphy- where CL refers to clearance, V to volume of distribution,
lococcus epidermidis (14), non-specified Staphylococcus CWT to current weight and PMA to post-menstrual age.
sp. (10) and Staphylococcus aureus (3). Remaining sepsis The final PK covariate model estimates and results from
events were attributed to Staphylococcus hominis (2), the bootstrap model are outlined in Table 2. The final model
Staphylococcus capitis (1), Staphylococcus warneri (1), E. accounted for some of the heterogeneity within the data
coli (2), Enterococcus faecalis (1) and Enterococcus as the predicted concentration vs. weighted residual plot
faecium (1). Five of the cultures contained mixed growth illustrated (Fig. 2).
with bacterial species and/or fungal species being isolated.
Forty-five amikacin E-tests were performed. Staphylococ- Pharmacodynamic model
cus epidermidis amikacin MIC ranged from 1.5 to 12 mg/L
and for non-specified Staphylococcus sp., MIC ranged There were 35 confirmed septic episodes from 26 neonates.
from 0.75 to 3 mg/L. Only 26 isolates were confirmed as Univariate analysis of the variables showed there was no
causative septic agents, 14 of which were from episodes of influence from amikacin AUC, sex, post-natal age, amikacin
treatment failure. peak or amikacin MIC upon treatment failure when included
in the logistic regression model as independent variables.
Pharmacokinetic model Other than amikacin peak/MIC ratio, there were no inde-
pendent predictors of treatment failure (Table 3). In those
The base PK model estimated amikacin clearance as with culture-proven sepsis, the relative risk (95% CI) for
0.069 L/h (0.486 SE %, 0.0590.078 95% CI) and volume treatment failure was 2.86 (1.575.19) for amikacin peak/
of distribution as 0.63 L (3.42 SE %, 0.560.69 95% CI), MIC <6 and 2.25 (0.955.34) for amikacin peak/MIC <8.

Table 1 Demographic summary of the amikacin study population, presented as median (range)

Characteristics of patients (n, %) Septic (n=31, 38.75%) Non-septic (n=49, 61.25%) All (n=80, 100%)

Gestational age (weeks) 27 (2434) 28 (2441) 28 (2441)

Post-natal age (days) 9 (364) 9 (331) 9 (364)
Post-menstrual age (weeks) 28.8 (24.741.1) 30.0 (24.744.0) 29.43 (24.744.0)
Birth weight (kg) 0.93 (0.582.1) 1.02 (0.44.4) 0.97 (0.444.4)
Current weight (kg) 0.97 (0.633.51) 1.20 (0.454.43) 1.03 (0.454.43)
Male sex (n, %) 23 (74%) 23 (46%) 46 (57%)
Apgar score at 1 min 7 (19) 7 (19) 7 (19)
Apgar score at 5 min 9 (210) 10 (510) 9 (210)
Serum creatinine (mol/L) 58 (42111) 59 (20111) 58.5 (20111)
C-reactive protein maximum (mg/L) 30 (0.5239) 1 (0.576) 2 (0.5239)
Eur J Clin Pharmacol (2009) 65:705713 709

5 5

3 3

1 1
WRES

WRES
-1 -1

-3 -3

-5 -5

25 28 31 34 37 40 43 0 8 16 24 32 40 48 56 64 72 80 88
PMA (weeks) PNA (days)
a) WRES vs. PMA base model b) WRES vs. PNA base model

5 5

3 3

1 1

WRES
WRES

-1 -1

-3 -3

-5 -5

25 28 31 34 37 40 43 0 8 16 24 32 40 48 56 64 72 80 88
PMA (weeks) PNA (days)
c) WRES vs. PMA final model d) WRES vs. PNA final model
Fig. 1 Diagnostic plots comparing the base and final models. WRES Weighted residual, PMA post-menstrual age, PNA post-natal age

Simulations of dosing regimens trough concentration below 5 mg/L. The present study
revised the target concentrations using the results of the PD
The existing amikacin dosing regimen recommended target analysis. The optimum target for Cmax was adjusted from a
concentrations for peaks between 20 and 30 mg/L with a peak of 2030 mg/L to 2435 mg/L (which was considered,

Table 2 NONMEM estimates of population pharmacokinetic parameters

Parameter Final model Bootstrap (n=1,000)

Estimates 95% CI Median 95% CI

Clearance (L/h) 0.23 0.173 to 0.287 0.23 0.173 to 0.306

CWT (3) 0.691 0.307 to 1.08 0.687 0.073 to 0.972
PMA (4) 3.23 1.82 to 4.64 3.23 1.805 to 4.955
Volume of distribution (L) 0.957 0.883 to 1.03 0.965 0.8725 to 1.03
CWT (5) 0.89 0.783 to 0.997 0.9 0.76 to 0.98
BSV-CL (%) 4.6 0.457 to 9.66 4.5 0.209 to 9.675
BSV-V (%) 0.446 0.181 to 1.07 0.477 0.03 to 1.015
RUV-CV (%) 3.8 1.88 to 5.72 4.0 1.68 to 6.44
RUV-SD (%) 2.42 0.44 to 4.4 2.09 0.56 to 3.88

CWT Current weight, PMA post-menstrual age, 95% CI confidence interval calculated from parameter standard errors, V volume of distribution,
CL clearance, BSV-CL between-subject variability related to clearance, BSV-V between-subject variability related to volume of distribution, RUV-
CV coefficient of variation of residual error (proportional error), RUV-SD coefficient of variation of residual error (additive error)
710 Eur J Clin Pharmacol (2009) 65:705713

56 because of the potential for toxicity, to be an acceptable

upper limit) [4, 6]. The upper and lower bounds for AUC24
Predicted concentrations (mg/L)

49
were 130 and 590 mgh/L, which were taken from the range
42 obtained by the univariate logistic regression for treatment
35
failure. The bounds for weight were 0.46.0 kg, and post-
menstrual age was adjusted to fit the required categories.
28
Table 4 outlines an example of percentage of success
21 achieved with simulated amikacin dosing for neonates with
a post-menstrual age 2327 weeks. A dosing regimen with
14
a dose range of 1415 mg/kg and dosing interval of 24
7 36 h based on post-menstrual age resulted in a high rate of
0 success for both AUC24 and Cmax on days 1 and 2 (Table 5).
0 7 14 21 28 35 42 49 56 Using this regimen, a serum amikacin concentration test
Observed concentrations (mg/L) could be performed at 24 h post-dose in order to detect
a) DV vs. PRED final model potential toxicity, for example using a criteria of trough
concentration <2 mg/L. If necessary, the dosing interval
could then be increased in the presence of unacceptably
5
high amikacin levels at 24 h post-dose.
3

Discussion
WRES

The principal findings of the present study were that the

-1
main determinants of amikacin clearance were current
weight and post-menstrual age, the main determinant of
-3
volume of distribution was current weight and the main
predictor of TF was peak/MIC ratio <8. Although the PK
-5
findings are similar to those previously reported, the present
0 9 18 27 36 45 54 study offers an extension of the PK modelling with
Predicted concentrations (mg/L) application of PD data in the simulation of a new dosing
regimen.
b) WRES vs. PRED final model
Overall, few population PK studies have been conducted
to describe the PK of amikacin in neonates [1416].
Published data on the use of amikacin are even more
5 limited for neonates with birth weights below 1.5 kg and
gestational age <30 weeks [17]. The BSV and residual
3
variability reported in the present study are lower than those
WRES

1 reported in previous studies [18]. The present study

included more extremely premature neonates (gestational
-1 age <28 weeks, 57.5%) and extremely low-birth-weight
infants (<1000 g, 55%) than previously published studies
-3
investigating amikacin PK in neonates. Consideration was
-5
given to investigating the potential effect of shrinkage on
the empirical Bayes estimates [19], but this was not
considered to explain the lower observed variability
0.0 3.2 6.4 9.6 12.8 16.0 19.2 22.4 25.6 28.8 32.0
determined in the final model of this study.
Time (hours) post dose
There is a wide BSV in relation to the PK of amikacin
c) WRES vs. time (hours) final model [5], and this makes it difficult to achieve a safe and
Fig. 2 Pharmacokinetic model: diagnostic plots for final covariate
effective dosing regimen particularly for extremely preterm
model. DV Dependent variable, PRED predicted concentration, WRES neonates [20, 21]. Differences in body composition and the
weighted residual renal immaturity associated with neonates contribute to the
variability of the PK parameters, particularly clearance [20].
Other variables that have been reported to influence
Eur J Clin Pharmacol (2009) 65:705713 711

Table 3 Univariate logistic

regression for treatment failure Exposure variable Number Median (range) or n (%) OR (95% CI)

MIC 25 3 (012) 1.30 (0.851.98)

Amikacin peak concentration 31 27.7 (17.136.8) 0.88 (0.741.04)
Current weight (kg) 35 1.0 (0.633.51) 1.35 (0.404.50)
C-reactive protein maximum (mg/L) 35 57 (0.5239) 1.01 (0.961.07)
AUC (mgh/L) 35 267.49 (133.39589.62) 1.01 (1.001.01)
OR Odds ratio, AUC area under Male sex 35 21 male (60) 0.42 (0.101.80)
the concentration-time curve,, Post-menstrual age (weeks) 35 29.57 (2541.14) 0.97 (0.781.21)
Cmax maximum peak concentra-
tion, MIC minimum inhibitory Amikacin peak concentration and MIC 22 10.8 (2.371.6) 1.02 (0.961.08)
concentration Cmax and MIC 24 9.2 (1.9162.9) 1.01 (0.981.05)
a
Values are relative risk instead Amikacin peak/MIC <6 22 2 (9%) 2.86 (1.575.19)a
of odds ratio (as predicts success Amikacin peak/MIC <8 22 4 (18%) 2.25 (0.955.34)a
perfectly)

variability include renal function, gestational age, post-natal development but did not provide the best minimisation to
age, haematocrit, fever and lean body weight [3]. the OFV. The study did not include neonates <72 h post-
The present study showed post-menstrual age to have the natal age as they were not treated with amikacin. As the
greatest significant effect on clearance. Post-menstrual age present study only included neonates of more than 72 h of
is suggested to be a predictor of amikacin clearance because age, the rapid changes associated with renal function in
it predicts the time course of development of glomerular neonates <72 h were not applicable to the model.
filtration rate (GFR) [22]. GFR matures during infancy and The effects of gestational age and post-natal age when
approaches an adult rate (6 L/h per 70 kg) by 6 months included as covariates in models are reported to provide
post-natal age [23]. Post-natal age and gestational age were disproportionate effects, and use of them as separate
investigated as separate covariates in the PK model during covariates may not be informative for renal clearance in

Table 4 Example of percentage of success achieved with simulated amikacin dosing for PMA 2327 weeks

Run PMA (weeks) Dose (mg/kg) Interval (h) AUC24 success AUC24 success Cmax success Cmax success
% day 1 % day 2 % day 1 % day 2

1 2327 19 48 99 99 24 9
2 2327 18 48 99 99 80 53
3 2327 17 48 99 99 94 81
4 2327 16 48 99 99 99 96
5 2327 15 48 97 98 98 99
6 2327 14 48 94 95 84 97
7 2327 13 48 91 92 47 75
8 2327 19 36 99 99 48 10
9 2327 18 36 99 99 76 30
10 2327 17 36 99 99 93 61
11 2327 16 36 99 99 98 85
12 2327 15 36 99 99 98 96a
13 2327 14 36 99 99 83 98
14 2327 13 36 99 99 50 90
15 2327 17 24 99 91 93 17
16 2327 16 24 99 95 99 44
17 2327 15 24 99 97 97 72
18 2327 14 24 99 99 88 93
19 2327 13 24 99 99 52 97

AUC24 Area under the concentration-time curve for 24 h, Cmax maximum peak concentration
a
Highest percentage success without the expense of prolonging dosing interval
712 Eur J Clin Pharmacol (2009) 65:705713

Table 5 Proposed dosing regimen based on achieving optimal AUC24 and Cmax

PMA (weeks) Dose (mg/kg) Interval (h) AUC24 success % AUC24 success % Cmax success % Cmax success %
day 1 day 2 day 1 day 2

28 15 36 99 99 98 97
2932 14 24 99 99 96 97
3336 14 24 97 97 99 98
37 15 24 78 76 95 93

Note: Dosing may be modified by measuring serum amikacin concentration at 24 h post-dose. Dosing regimen does not apply to neonates of
<72 h post-natal age
PMA Post-menstrual age, AUC24 area under the concentration-time curve for 24 h, Cmax maximum peak concentration

neonates >72 h age. This is stated to be particularly evident neonates cannot respond to haemodynamic adjustments as
beyond the initial neonatal period, and therefore it may be effectively as full-term neonates [31]. This suggests that
more appropriate to use post-menstrual age [21, 24]. The amikacin dosage regimens should be individualised espe-
PK base model in the present study estimated clearance of cially in the 34- to 38-week gestational age group [5] by
0.069 L/h (1.15 mL/min) in neonates with a gestational age adjusting the dose according to body weight and the dosing
of 2441 weeks. Published PK models have previously interval according to both body weight and post-menstrual age.
reported amikacin clearance as 0.75 and 0.87 mL kg1 min1 Results from the present study indicated the prime
in neonates of less than 31 weeks gestation [8, 25]. In determinant of treatment failure was a peak/MIC <8. The
neonates less than 72 h old, with the majority comedicated ability to achieve a therapeutic maximum serum concentra-
with non-steroidal anti-inflammatory drugs, amikacin clear- tion (Cmax) can be critical to patient survival because of a
ance has been reported as reduced, at 0.60 mL kg1 min1 significant decrease in the rate of mortality due to infection
for gestational age 2830 weeks and 0.44 mL kg1 min1 in critically ill patients [3, 21, 3234]. Hence, extended-
for gestational age <28 weeks [23, 25]. interval dosing regimens are recommended as they are
The primary determinant of amikacin volume of distri- expected to result in higher Cmax and greater efficacy [5, 8,
bution in the present study was current weight. Previous 13, 20, 21, 23, 33]. However, there is also a risk that
studies found the covariates influencing V in neonates to be extending the dosage interval beyond 24 h might provide
sepsis, body water and post-natal age [14, 26]. Patients with an opportunity for bacterial re-growth.
severe sepsis may have substantial changes in the extra- There were various limitations associated with the
cellular fluid volume which may alter the volume of present study and this was reflected in the results of the
distribution of many drugs [27]. Based on their relative PK model, PD models, the simulations and the proposed
higher water content, preterm neonates already have a higher dosing regimens for amikacin in this neonatal population.
volume of distribution for hydrophilic drugs. Therefore, a There were a limited number of neonates used for the PK
relatively higher dose per kilogram of amikacin is deemed and PD analyses. However, numbers will always be limited
necessary in more-preterm neonates to achieve a sufficient in any prospective observational study that identifies
concentration at the site of infection [20]. adverse outcomes (e.g. treatment failure) because once a
Although previous studies report substantial changes in problem is identified clinicians will take immediate
volume of distribution during the first weeks of life, the corrective action. In addition, the dataset consisted of 55%
present study included only neonates more than 72 h of age extremely low-birth-weight infants (< 1,000 g) and 57.5%
and hence might not have detected these changes [5, 8, 21, extremely premature neonates (gestational age <28 week),
25, 28]. An increase in volume of distribution associated providing a possible bias in the PK model. The number of
with gestational age is suggested to be related to an increase older neonates was limited as the population dataset
in lean body mass particularly around 36 weeks gestational contained only 10 (12.5%) neonates with gestational age
age rather than an increase in fat body mass. In neonates, 34 weeks.
anatomical and functional immaturity of the kidney limits
glomerular and tubular functional capacity. Renal matura- Acknowledgements The authors would like to acknowledge Pro-
tion has been shown to be related to gestational age [29]. fessor Stephen Duffull, School of Pharmacy, University of Otago,
Dunedin, New Zealand, for his assistance with the PK modelling and
Changes in GFR in neonates are related to nephrogenesis, the simulations. Catherine Sherwin is supported by a Freemasons
which is not complete until around 34 weeks gestational Postgraduate Fellowship in Paediatrics and Child Health from the
age [30]. Incomplete nephrogenesis means that premature Freemasons of New Zealand.
Eur J Clin Pharmacol (2009) 65:705713 713

References 18. Berger A, Kretzer V, Gludovatz P, Heinze G, Haiden N, Pollak A

(2004) Evaluation of an amikacin loading dose for nosocomial
infections in very low birthweight infants. Acta Paediatrica
1. Clapp DW (2006) Developmental regulation of the immune 93:356360
system. Semin Perinatol 30:6972 19. Savic RM, Karlsson MO (2007) Importance of shrinkage in
2. Petersen PO, Wells TG, Kearns GL (1991) Amikacin dosing in empirical bayes estimates for diagnostics and estimation: prob-
neonates: evaluation of a dosing chart based on population lems and solutions. Abstracts of the Annual Meeting of the
pharmacokinetic data. Dev Pharmacol Ther 16:203211 Population Approach Group in Europe, p. 16
3. Zaske DE, Strate RG, Kohls PR (1991) Amikacin pharmaco- 20. Allegaert K, Cossey V, Langhendries JP, Naulaers G, Vanhole C,
kineticswide interpatient variation in 98 patients. J Clin Devlieger H, Van Overmeire B (2004) Effects of co-
Pharmacol 31:158163 administration of ibuprofen-lysine on the pharmacokinetics of
4. Bristol-Meyers S (2001) Amikin. Bristol-Myers Squibb Pharma- amikacin in preterm infants during the first days of life. Biol
ceuticals, Noble Park, Australia Neonate 86:207211
5. Bleyzac N, Varnier V, Labaune JM, Corvaisier S, Maire P, 21. Treluyer JM, Merle Y, Tonnelier S, Rey E, Pons G (2002)
Jelliffe RW, Putet G, Aulagner G (2001) Population pharmaco- Nonparametric population pharmacokinetic analysis of amikacin
kinetics of amikacin at birth and interindividual variability in in neonates, infants, and children. Antimicrob Agents Chemother
renal maturation. Eur J Clin Pharmacol 57:499504 46:13811387
6. Young TE, Mangum B (2002) Neofax, 15 ed. Acorn Publishing, 22. Koren G, James A, Perlman M (1985) Simple method for the
Raleigh estimation of glomerular filtration rate by gentamicin pharmaco-
7. Craig WA (1998) Pharmacokinetic/pharmacodynamic parameters: kinetics during routine drug monitoring in the newborn. Clin
rationale for antibacterial dosing of mice and men. Clin Infect Dis Pharmacol Ther 38:680685
26:112 23. Allegaert K, Anderson BJ, Cossey V, Holford NHG (2005)
8. Langhendries JP, Battisti O, Bertrand JM, Francois A, Kalenga M, Limited predictability of amikacin clearance in extreme premature
Darimont J, Scalais E, Wallemacq P (1998) Adaptation in neonates at birth. Br J Clin Pharmacol 61:3948
neonatology of the once-daily concept of aminoglycoside admin- 24. Bueva A, Guignard JP (1994) Renal function in preterm neonates.
istration: evaluation of a dosing chart for amikacin in an intensive Pediatr Res 36:572577
care unit. Biol Neonate 74:351362 25. Kenyon CF, Knoppert DC, Lee SK, Vandenberghe HM, Chance
9. Cookson B, Tripp J, Leung T, Williams JD (1980) Evaluation of GW (1990) Amikacin pharmacokinetics and suggested dosage
amikacin dosage regimes in the low and very low birthweight modifications for the preterm infant. Antimicrob Agents Chemother
newborn. Infection 8(Suppl 3):239242 34:265268
10. Ette EL, Williams PJ, Kim YH, Lane JR, Liu MJ, Capparelli EV 26. Lingvall M, Reith DM, Broadbent R (2005) The effect of sepsis
(2003) Model appropriateness and population pharmacokinetic upon gentamicin pharmacokinetics in neonates. Br J Clin
modeling. J Clin Pharmacol 43:610623 Pharmacol 59:5461
11. Pagano M, Gauvreau K (1993) Principles of biostatistics. 27. Mann HJ, Fuhs DW, Awang R, Ndemo FA, Cerra FB (1987)
Wadsworth Publishing, Belmont, CA Altered aminoglycoside pharmacokinetics in critically ill patients
12. Kotze A, Bartel PR, Sommers DK (1999) Once versus twice daily with sepsis. Clin Pharm 6:148153
amikacin in neonates: prospective study on toxicity. J Paediatr 28. Padovani EM, Pistolesi C, Fanos V, Messori A, Martini N (1993)
Child Health 35:283286 Pharmacokinetics of amikacin in neonates. Dev Pharmacol Ther
13. Langhendries JP, Battisti O, Bertrand JM, Francois A, Darimont J, 20:167173
Ibrahim S, Tulkens PM, Bernard A, Buchet JP, Scalais E (1993) 29. Besunder JB, Reed MD, Blumer JL (1988) Principles of drug
Once-a-day administration of amikacin in neonates: assessment biodisposition in the neonate. A critical evaluation of the
of nephrotoxicity and ototoxicity. Dev Pharmacol Ther 20:220 pharmacokinetic-pharmacodynamic interface (part ii). Clin
230 Pharmacokinet 14:261286
14. Botha FJH, van der Bijl P, Seifart HI, Parkin DP (1996) 30. van den Anker JN (1996) Pharmacokinetics and renal function in
Fluctuation of the volume of distribution of amikacin and its preterm infants. Acta Paediatr 85:13931399
effect on once-daily dosage and clearance in a seriously ill patient. 31. Aperia A, Broberger O, Elinder G, Herin P, Zetterstrom R (1981)
Intensive Care Med 22:443446 Postnatal development of renal function in pre-term and full-term
15. Assael BM, Parini R, Rusconi F, Cavanna G (1982) Influence of infants. Acta Paediatr Scand 70:183187
intrauterine maturation on the pharmacokinetics of amikacin in the 32. Moore RD, Smith CR, Lietman PS (1984) Association of
neonatal period. Pediatr Res 16:810815 aminoglycosides plasma levels with therapeutic outcome in
16. Botha JH, du Preez M, Miller R, Adhikari M (1998) Determina- gram-negative pneumonia. Am J Med 77:657662
tion of population pharmacokinetic parameters for amikacin in 33. Moore RD, Lietman PS, Smith CR (1987) Clinical response to
neonates using mixed-effect models. Eur J Clin Pharmacol aminoglycoside therapy: importance of the ratio of peak concen-
53:337341 tration to minimal inhibitory concentration. J Infect Dis 155:9399
17. Kenyon CF, Knoppert DC, Lee SK, Vandenberghe HM, Chance 34. Noone P, Parsons TM, Pattison JR, Slack RC, Garfield-Davies D,
GW (1990) Amikacin pharmacokinetics and suggested dosage Hughes K (1974) Experience in monitoring gentamicin therapy
modifications for the preterm infant. Antimicrob Agents Chemother during treatment of serious gram-negative sepsis. Br Med J
34:265268 1:477481